粒子濾波器重采樣算法的研究與改進(jìn).pdf

1、JUlIIIEIIIIHIIIIrlIII]IIIfI]IJIJlJUfY1437794分類號：UDCo密級：編號：工學(xué)碩士學(xué)位論文粒子濾波器重采樣算法的研究與改進(jìn)碩士研究生：王萌指導(dǎo)教師：馮馳教授學(xué)位級別：工學(xué)碩士學(xué)科、專業(yè)：信號與信息處理所在單位：哈爾濱工程大學(xué)論文提交日期：2008年3月論文答辯日期：2008年3月學(xué)位授予單位：哈爾濱工程大學(xué)哈爾濱‘I：稃人學(xué)碩十學(xué)位論文ABSTRACTParticlefilterisanalgo

2、rithmbasedonMonteCarloandrecursiveBayesianestimationItsbasicideaistouseseriesofweightedsamplestoapproximateposteriorprobabilitydensitydistributioninthestatespaceandtopresentintegraloperationbysamplemeanvalueInprinciple，p

3、articlefiltercanrealizeanystateestimation，especiallyshowsexcellentperformanceinnon—linear／non—GaussianproblemswhileKalmanfilterandextendedKalmanfilterloseefficiencyParticlefilterhasbeenwidelyusedinthefieldssuchastargettr

4、acking，visiontracking，faultdiagnosis，navigationpositionandradiocommunicationetcResamplingalgorithmisanimportantstepofpaniclefilter，andisoneofimportantmethodstosolvedegeneracyproblemoftheparticlefilterThisthesisfirstlystu

5、diesparticlefilteringalgorithmindepth，introducesrealizationtheoryandstepsofthealgorithmindetail，andbrieflydiscusseschoicesofimportancedensityfunctionandresamplingalgorithm；thenthisthesistheoreticallyanalyzesfourclassicre

6、samplingalgorithmsincludingmultinomialresampling，residualresampling，strmifiedresamplingandsystematicresampling，frompointsofresamplingrealizationprinciple，resamplingqualityandcomputationcomplexity，andcomparesperformanceso

7、fthefourresamplingalgorithmsviaexperimentsimulationBasedonresidualresamplingandresidualsystematicresampling，thisthesispresentsanimprovedresidualresamplingalgorithm，themostdifferenceoftheimprovedalgorithmcomparedwithresid

8、ualresamplingalgorithmliesinthatthisimprovedalgorithmdoesnotcomputethepaniclecopiedtimesseparatelybutfirstlycomputesthesumoftheproductofpanicleweightsandthetotalparticlenumber，thenrounds，ensuringtheinvariabilityofthepart